Closed andrewgettelman closed 2 years ago
In MG4 I have added a max value for allowed ice number concentration for testing. This is a nametlist parameter in MG4. It is listed in MG3 as well, but I am not using it in MG3 yet. I was actually about to ask you Andrew if you want me to incorporate this into MG3 as well?
-- Trude
On Wed, Jan 20, 2021 at 8:58 PM Andrew Gettelman notifications@github.com wrote:
An issue highlighted by Zachary McGraw and Trude Storelvmo in Oslo:
'nimax' sets the maximum ice number. It was designed to include the Meyers ice nucleation, but was not modified to include the Hoose classical nucleation theory (Mixed phase ice nucleation). So it might limit the impact here:
https://github.com/ESCOMP/PUMAS/blob/97b13fde8e9f97c97324e66904af0b865d0269a6/micro_mg3_0.F90#L2425
Question: Do we need nimax anymore? At a minimum we should include the mixed phase ice nucleation (immersion + deposition + contact: frzimm, frzcnt, frzdep or the corresponding nnuccc, nnucct, nnudep) in nimax....
This will need some testing...
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/PUMAS/issues/20, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD24I5ISY76DDSR4WU6EYV3S26Q5NANCNFSM4WMC374Q .
-- Dr. Trude Eidhammer Project Scientist II Research Applications Laboratory (RAL) National Center for Atmospheric Research (NCAR) P.O. Box 3000 Boulder, CO 80307-3000, USA
Thanks for the reply Trude. nimax I think is variable in the code: it is set to the INP basically, and the issue is it probably does not include the new Hoose mixed phase INP. But it will also cap ice multiplication as well. I'm tagging @hmorrison100 on this thread to see if he has an opinion. It's going to need some test simulations (probably just with MG3) to see what the impact is. I may try it in the IFS version to see if that does anything.
Update from Zachary McGraw in Oslo:
We think nimax makes sense except that its current setting is preventing ice sources in mixed-phase clouds from acting as intended.
We've tried two separate methods of altering the nimax implementation: i) turning nimax off entirely for temperatures warmer than cirrus (eg T>-40C)
ii) adding in the Hoose terms to resemble the treatment of Meyers (Hoose outputs rates rather than number densities so nnuccclcldmdeltat, etc).
Personally speaking, we liked the second method more because a) it's most in line with previous model versions that worked for our purpose (mixed-phase cloud FB & ACI research) and b) we didn't want to modify the model more than necessary for our research. However, I can confirm as you mention that secondary ice multiplication is also suppressed by nimax (and becomes quite strong with nimax disabled in its temperature range).
FYI: I did a quick test of removing nimax in the ECMWF-IFS single column case over the S. Ocean, and it did not have much impact on ice number in this case. Might have to run it globally.....
I'm actually a little concerned with "nimax". This is an ancient piece of code (going back to the very first version of MG 15 years ago) to avoid excessive nucleation with the long time step. But this is for nucleation on IN, and it should not be applied to ice multiplication (which was added to MG only later). Homogeneous nucleation is not included in the limit, since it's calculated below these lines where the limiter is applied:
if (do_cldice .and.
nitend(i,k).gt.0._r8.and.ni(i,k)+nitend(i,k)*deltat.gt.nimax(i,k)) then nitend(i,k)=max(0._r8,(nimax(i,k)-ni(i,k))/deltat) end if
What you could to is add ice multiplication/secondary ice production to nitend only after these lines, not before them. Or better, just remove the limiter entirely -- what it's intended to fix may not be a problem anymore anyway with the shorter time step now. Good to try a global test, and if it looks ok my suggestion would be remove this.
Hope this helps.
Cheers, Hugh
On Thu, Jan 21, 2021 at 5:41 PM Andrew Gettelman notifications@github.com wrote:
FYI: I did a quick test of removing nimax in the ECMWF-IFS single column case over the S. Ocean, and it did not have much impact on ice number in this case. Might have to run it globally.....
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/PUMAS/issues/20#issuecomment-765034659, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJDVIE7QE43Q77TT7U6DYO3S3DCT5ANCNFSM4WMC374Q .
I agree the ideal would be for the 'nimax' limit to be removed, assuming this doesn't enable the affected processes to cause new issues. Removing the limit frees up 3 processes to watch: i) local ice nucleation in mixed-phase clouds via Hoose ('nnucct','nnuccc','nnudep'), ii) secondary ice production ('nsacwi'), and iii) conversion from falling rain to standing ice ('nnuccri').
Turning off nimax won't directly affect the remaining ice number sources, which are either a) already factored into nimax (cirrus nucleation) or located outside the microphysics loop containing nimax (eg homog. freezing, as mentioned, and ice sedimentation from above).
Zachary and I tested removing NIMAX at temperatures above -38C in the Arctic to see the effect on supercooled liquid fraction (SLF) and cloud phase (see figures below). These experiments were done with atmosphere-only in NorESM2 (CAM6, micro_mg2) and we only looked at the Arctic, so their usefulness is limited. But we did see that removing NIMAX significantly lowers SLF and the liquid cloud fraction.
I am not as familiar with micro_mg2 as Zachary, but these were my thoughts on the code: It is my understanding that when the ice number tendency is produced diagnostically and ice nucleation does not modify the aerosol fields, the "same" aerosols are in effect able to nucleate additional ice at each timestep. NIMAX prevents this, but the limit it imposes is somewhat arbitrary and it does not represent a physical process.
Separately, while the implementation of NIMAX mostly prevents the Hoose nucleation scheme from adding to ice number, the contribution to the ice mass tendency terms are unaffected and can increase the ice crystal size.
Jonah
I did a simulation where I turn off nimax in CAM6 (cam6_2_32). This does increase ice and decrease liquid quite a bit. Bottom line: nimax has a large impact. It would need some tuning or the more nuanced treatment as suggested.
When I was investigating the MG4 low/high biases in cloud radiative forcing, one thing I noticed was the impact in restricting the number of ice crystals (note, I did not restrict any ice nucleation/freezing), but the total amount to ice number concentration.
On Mon, Jan 25, 2021 at 4:19 PM Andrew Gettelman notifications@github.com wrote:
I did a simulation where I turn off nimax in CAM6 (cam6_2_32). This does increase ice and decrease liquid quite a bit. Bottom line: nimax has a large impact. It would need some tuning or the more nuanced treatment as suggested.
[image: zm_nimax_diff] https://user-images.githubusercontent.com/9616874/105778142-f6454f00-5f28-11eb-9a8e-384e977ea6ae.png
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/PUMAS/issues/20#issuecomment-767175201, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD24I5IYOGROKSL6QPEF7QDS3X37DANCNFSM4WMC374Q .
-- Dr. Trude Eidhammer Project Scientist II Research Applications Laboratory (RAL) National Center for Atmospheric Research (NCAR) P.O. Box 3000 Boulder, CO 80307-3000, USA
Just to keep up the thread: I have a suggested fix for this (NIMAX removed plus an ice number limiter at the end) which will come in when I get a chance to bring in a bunch more PUMAS changes I have (after the MG3-GPU and PPE dust settles).
An issue highlighted by Zachary McGraw and Trude Storelvmo in Oslo:
'nimax' sets the maximum ice number. It was designed to include the Meyers ice nucleation, but was not modified to include the Hoose classical nucleation theory (Mixed phase ice nucleation). So it might limit the impact here:
https://github.com/ESCOMP/PUMAS/blob/97b13fde8e9f97c97324e66904af0b865d0269a6/micro_mg3_0.F90#L2425
Question: Do we need nimax anymore? At a minimum we should include the mixed phase ice nucleation (immersion + deposition + contact: frzimm, frzcnt, frzdep or the corresponding nnuccc, nnucct, nnudep) in nimax....
This will need some testing...